Standard Report on Methods and Quality. for the Labour Force Survey (LFS)

Transcription

1 Standard Report on Methods and Quality for the Labour Force Survey (LFS) This documentation applies to the current reporting period: 2017 Quarter 4 Last edited: <13 March 2018> CENTRAL STATISTICS OFFICE Skehard Road, Cork

2 Table of Contents 1 Overview 2 General Information 2.1 Statistical Category 2.2 Area of Activity 2.3 Organisational Unit Responsible, Persons to Contact 2.4 Objectives and Purpose; History 2.5 Periodicity 2.6 Client 2.7 Users 2.8 Legal basis 3 Statistical Concepts, Methods 3.1 Subject of the Statistics 3.2 Units of Observation/Collection Units/Units of Presentation 3.3 Data Sources 3.4 Reporting Unit/Respondents 3.5 Type of Survey/Process 3.6 Characteristics of the Sample/Process Population and Sampling Frame Sampling Design 3.7 Survey Technique/Data Transfer 3.8 Questionnaire (including explanations) 3.9 Participation in the Survey 3.10 Characteristics of the Survey/Process and its Results 3.11 Classifications used 3.12 Regional Breakdown of Results 4 Production of the Statistics, Data Processing, Quality Assurance 4.1 Data Capture 4.2 Coding 4.3 Data Editing 4.4 Imputation (for Non-Response or Incomplete Data Sets) 4.5 Census 2016 Population revisions 4.6 Backcasting of QNHS data series 4.7 Grossing and Weighting 4.8 Computation of Outputs, Estimation Methods Used 4.9 Other Quality Assurance Techniques Used 5 Quality 5.1 Relevance 5.2 Accuracy and Reliability Sampling Effects, Representativity Non-Sampling Effects Quality of the Data Sources used Register Coverage Non-response (Unit and Item) 2

3 Measurement Errors Processing Errors Model-related Effects 5.3 Timeliness and Punctuality Provisional Results Final Results 5.4 Coherence 5.5 Comparability 5.6 Accessibility and Clarity Assistance to Users, Special Analyses Revisions Publications Releases, Regular Publications Statistical Reports Internet Confidentiality 6 Additional documentation and publications 3

4 1 Overview The Labour Force Survey (LFS) replaced the Quarterly National Household Survey (QNHS) from Q The survey meets the requirements of Council Regulation (EC) No.577/98, adopted in March 1998, which requires the introduction of quarterly labour force surveys in EU Member States. Outputs are primarily used to produce quarterly labour force, employment and unemployment estimates for Ireland. The LFS may also be used as a means of collecting data on important social topics, included as modules in the survey for EU requirements. Outputs are also supplied to other internal users such as National Accounts, Tourism and Earnings. The LFS is conducted using mixed mode data collection with the introduction of Computer Assisted Telephone Interviewing (CATI). As with the QNHS, information is collected from each sample household over 5 successive quarters or waves. However in the LFS, the first interview is conducted by a team of face-to-face interviewers using Computer Assisted Personal Interviewing (CAPI). The four follow-up interviews are conducted using CATI from a dedicated call centre, where householders have agreed to conduct a telephone interview, and are conducted using face-to-face interviews where householders have not agreed to conduct a telephone interview. Information is collected continuously throughout the year from households surveyed each week in each quarter. To account for the additional attrition as a result of the introduction of mixed mode data collection, the LFS sample is being increased incrementally from Q An additional 1,300 households have been included in Wave 1 and results in a total quarterly sample of 28,600 households in Q As the sample is being increased by 1,300 households per wave, the final design sample will increase to 32,500 households by Q The actual achieved sample varies over time depending on the level of response. The achieved sample can be seen in section

5 2 General Information 2.1 Statistical Category Primary Statistical survey 2.2 Area of Activity Labour Market 2.3 Organisational Unit Responsible, Persons to Contact The relevant sections are part of the Social and Demographic Statistics Directorate. The work of LFS section is largely divided into two areas a Household Survey Collection Unit (HSCU) and a Labour Market Analysis unit. Each unit is headed by a Senior Statistician who also has responsibility for other survey areas. Labour Market Analysis queries Jim Dalton Tel: labour@cso.ie Martina O Callaghan Tel: labour@cso.ie Household Survey Collection Unit queries Kevin Healy Tel: qnhs@cso.ie 2.4 Objectives and Purpose; History The LFS has replaced the QNHS from Q The primary purpose of the survey is to collect information on the Irish labour market (collected via what is known as the CORE questionnaire. It may also be used to pilot additional labour market related questions. 2.5 Periodicity Information is collected continuously throughout the year. The sample is designed to be representative on a quarterly basis with a target of 2,200 households to be surveyed each week. This resulted in a total design sample of 28,600 households for Q The reference quarters for survey results are: Q1- January to March, Q2- April to June, Q3- July to September and Q4- October to December. (i.e. calendar quarters). The original QNHS changed from seasonal quarters to calendar quarters in Q Prior to that, the QNHS operated on a seasonal quarter basis. (Q1- December to February, Q2- March to May, Q3- June to August, Q4- September to November) For further information on the changeover to calendar quarters, see the supplementary note linked below. 5

6 Results previously published on a seasonal quarter basis were reissued on a calendar quarter basis for the periods Q to Q Client Each Member State in the EU must undertake a Labour Force Survey (LFS) to provide information on key labour market indicators. The LFS is the Irish implementation of the EU-LFS. The LFS also provides important inputs to national policy makers. 2.7 Users European Union/Eurostat Government departments (Department of the Taoiseach, Department of Finance, Department of Business, Enterprise and Innovation, Department of Employment Affairs and Social Protection, Department of Education and Skills etc) SOLAS National Skills Training Agency Other research centres and universities involved in labour market research National media The general public 2.8 Legal basis The LFS is carried out under EU Council Regulation No 577/98. There are also a number of implementing regulations which outline various aspects of the survey, how data should be coded for transmitting to Eurostat, the topics for pan European ad-hoc modules etc. basis While Ireland as a Member State is obliged to undertake the LFS, participation in the survey is voluntary for respondents. 6

7 3 Statistical Concepts, Methods 3.1 Subject of the Statistics Labour market statistics. 3.2 Units of Observation/Collection Units/Units of Presentation The survey population is individuals living in private households. It therefore excludes individuals living in institutions or communal accommodation and persons of no fixed abode. The collection units are households containing at least one individual aged 15 years or over for whom it is the main residence. Information is collected on each individual within a surveyed household. The main units of presentation are: Demographic variables such as Sex, Age, Nationality Other personal characteristics such as Regional classification NUTS 2 & 3 (Nomenclature of Territorial Units), Highest level of education attained etc. International Labour Office labour force classification (ILO Status) Industrial activity classification NACE Rev.2 (Nomenclature des Activités de la Communauté Européenne) Occupation Employment status While the above are the primary presentation units for regular publication the LFS can present data according to a wide variety of classifications based on the comprehensive range of questions asked in the survey. Such analysis is often provided on an ad hoc basis following user requests. 3.3 Data Sources Information is collected from individuals in households. 3.4 Reporting Unit/Respondents All usual residents 1 ' in responding households are surveyed. Where a particular individual is not available for interview, information can be provided by another member of the household in most circumstances via a proxy interview. A proxy interview refers to data which is collected from another member of the household due to the unavailability of the specific respondent at the time of the interview. 1 A person is defined as a "Usual Resident" of a private household if he or she (i) Lives regularly at the dwelling in question and (ii) Shares the main living accommodation (i.e. kitchen, living room or bathroom) with the other members of the household. 7

8 3.5 Type of Survey/Process Sample survey. 3.6 Characteristics of the Sample/Process Population and Sampling Frame The reference population is all individuals living in private households in Ireland. It therefore excludes persons with no usual address or those with a usual residence in a public institution, such as hospitals, nursing homes etc. All usual residents in each household are included. The sampling frame is all private households in Ireland. Beginning in Q a new sample for the LFS based on the 2011 Census of Population was introduced incrementally and this sample was fully in effect as of Q It is important to note however that there is no overlap in sample between the QNHS and the LFS Sample Design Sample size and design was initially determined after consultation with relevant experts 2,3 who considered a mix of factors including cost, levels of precision of estimates etc. The LFS sampling approach is based on the review by Haase Pratschke. With this design, a two-stage sample design is used. The sample frame of households is clustered into blocks (small areas) with each block containing a minimum of 60 occupied households on the night of the 2011 Census of Population. The sample is stratified using administrative county and the Pobal HP (Haase and Pratschke) Deprivation Index. In the first stage 1,300 blocks are selected using Probability Proportional to Size (PPS) sampling and in the second stage 20 households are selected using Simple Random Sampling (SRS). To account for the additional attrition as a result of the introduction of mixed mode data collection, the LFS sample is being increased incrementally from Q An additional 1,300 households have been included in Wave 1 and results in a total quarterly sample of 28,600 households in Q As the sample is being increased by 1,300 households per wave, the final design sample will increase to 32,500 households by Q The actual achieved sample varies over time depending on the level of response. Households are asked to take part in the survey for five consecutive quarters and are then replaced by other households in the same block. Thus, one fifth of the households in the 2 Steel report, httphttp:// 3 CSO_Household_Surveys.pdf 8

9 survey are replaced each quarter and the LFS sample involves an overlap of 80% between consecutive quarters and 20% between the same quarters in consecutive years. The new LFS sample based on the 2011 Census of Population was introduced incrementally from Q and is fully effective from Q It is important to note that there is no overlap in sample between the QNHS and the LFS. 3.7 Survey Technique/Data Transfer The survey is carried out using mixed mode data collection. The survey is voluntary and participating households are surveyed over five successive quarters. The first interview is carried out with a CSO interviewer using CAPI on encrypted laptop computers and every night the data collected is transferred to the CSO using a secure encrypted data tunnel. The four follow-up interviews are conducted using CATI from a dedicated call centre, where householders have agreed to conduct a telephone interview, and are conducted using face-to-face interviews where householders have not agreed to conduct a telephone interview. Each night, the data collected that day is collated into a single zip file, encrypted and password protected. The file is then uploaded to a secure Secure File Transfer Protocal (SFTP) site at the dedicated call centre. This CSO then extracts this zip file to a secure CSO location for processing. 3.8 Questionnaire (including explanations) The LFS questionnaire contains approximately 200 questions on a range of topics including the respondents economic status (i.e. employed, unemployed, not in the labour force), industry of employment, nationality, employment status, occupation, education level, length of time unemployed etc. It can be noted that not all questions are asked of all respondents as questions are filtered based on the responses given by the interviewee. Additional questions on a particular topic may be included for modules which are run in individual quarters to meet EU requirements. The LFS core questionnaire is currently being updated and will be published as soon as possible. In the meantime, the QNHS core questionnaire (not too dissimilar to the LFS) and module questionnaires are available at the following link to the CSO website Participation in the Survey 9

10 Participation in the survey is voluntary Characteristics of the Survey/Process and its Results The survey is designed to produce Labour Force estimates for the State. The data which are published in the main quarterly release refer to the estimated number of persons aged 15 years or over in the state classified by various characteristics including economic status (e.g. employed, unemployed, not in the labour force) in accordance with the ILO (International Labour Organisation) definition. Three key rates are also produced from the survey: the participation rate (the proportion of all persons aged 15 or more who are in the Labour Force), the employment rate (the proportion of all persons aged who are in employment) and the unemployment rate (the proportion of all persons aged in the Labour Force who are unemployed. The calculation of the unemployment rate has been revised slightly as of Q to ensure coherence with Eurostat. Prior to this, the Unemployment Rate was calculated as the number of unemployed expressed as a percentage of the total labour force aged 15 and over. The change introduced limits the labour force to persons aged and this excludes a small number of persons aged 75 and over in employment from the total labour force used in the calculation. The overall impact of this change was minimal. Economic status is in turn classified by various characteristics such as age, sex, region, educational level, etc. Some of the key indicators produced are estimates of those in employment by occupation and industry and estimates of those unemployed by duration of unemployment. For full details of published indicators please see the latest LFS release on Classifications used There are a number of different classifications used in the LFS: The main classifications of economic activity are the standard ILO definitions of persons in employment, unemployment (which are summed to derive the labour force total) and persons not in labour force. Industry is published by the NACE Rev 2 classification. Occupation is primarily published using the UK SOC 2010 classification for national purposes and is also coded to ISCO-08 classification for EU purposes. For EU purposes, education details have been coded to ISCED 1997 (International Standard Classification of Education) up to Q4, 2013 and to ISCED 10

11 2011 thereafter. A national classification for Irish purposes is also available. Field of education is also published. Regional data is coded to NUTS3 as described in Section 3.12 Detailed data is collected through the interview to allow outputs to be produced according to all the relevant classifications Regional Breakdown of Results The regional classifications used are based on the NUTS (Nomenclature of Territorial Units) classification used by Eurostat. The NUTS3 regions correspond to the eight Regional Authorities established under the Local Government Act, 1991 (Regional Authorities) (Establishment) Order, 1993, which came into operation on 1 January The NUTS2 regions, which were proposed by Government and agreed by Eurostat in 1999, are groupings of the NUTS3 regions. The sample itself is designed to be representative at NUTS3 level although some results may be repressed depending on the number of observations in the particular cell. The composition of the regions is set out below: NUTS2: Border, Midlands and Western Southern and Eastern NUTS3: Border Cavan Dublin Dublin Donegal Dún Laoghaire-Rathdown Leitrim Fingal Louth South Dublin Monaghan Sligo Mid-East Kildare Meath Midlands Laoighis Wicklow Longford Offaly Mid-West Clare Westmeath Limerick City Limerick County West Galway City North Tipperary Galway County Mayo South-East Carlow Roscommon Kilkenny South Tipperary Waterford City Waterford County Wexford South-West Cork City Cork County Kerry The NUTS 3 boundaries were amended on 21 st of November 2016 under Regulation (EC) No 2066/2016 and will come into force from Q These will be reflected in the LFS 11

12 results from Q onwards. The changes resulting from the amendment are that Louth has moved from the Border to the Mid-East and what was formerly South Tipperary has moved from the South-East to the Mid-West. 4 Production of the Statistics, Data Processing, Quality Assurance 4.1 Data Capture The survey is carried out using mixed mode data collection. The survey is voluntary and participating households are surveyed over five successive quarters. The first interview is carried out by face-to-face interview with a CSO interviewer using CAPI. The four follow-up interviews are conducted using CATI from a dedicated call centre, where householders have agreed to conduct a telephone interview, and are conducted using face-to-face interviews where householders have not agreed to conduct a telephone interview. 4.2 Coding In the LFS, interviewers collect a detailed description of the enterprise and occupation from respondents. This information is then coded in-house at the CSO by an automated process which is reviewed by a small dedicated team of coding experts using a series of consistency checks. Field of education data is captured and coded in the field to the relevant classification (see Section 3.11) while the region of place of work is coded using the 34 administrative counties (see Sections 3.11 and 3.12). 4.3 Data Editing The majority of questions only allow answers to be entered to a limited set of predefined categories and therefore the number of edits required is limited. Questionnaire routing is used to ensure questions are only asked to relevant respondents e.g. unemployment questions are only asked to those who are unemployed. In addition, invalid responses are prevented at the point of capture where appropriate (for example a flag is raised to an interviewer if the date of birth captured which imply a derived age of more than 110) and this ensures that implausible data is prevented from being captured. The most significant data editing done is the quality checking review of the industry and occupation coding. 4.4 Imputation (for Non-Response or Incomplete Data Sets) No imputation for non-response currently takes place on the LFS, either for entirely missing households or missing data for particular individuals. Proxy interviews are allowed to obtain data for respondents who are not present in the house at time of interview. Up to 50% of interviews are proxy interviews where information has been provided by another resident of the household due to unavailability of the person in question. There are known issues with the quality of data for proxy responses for certain information in particular. For example while a proxy respondent may know the age of 12

13 other residents in the household, they may not know how long they have worked with their current employer (particularly in shared households where residents are not related). 4.5 Census 2016 Population revisions As the LFS (and previously the QNHS) is a sample survey, independent population estimates are required each quarter to provide a weighting basis for the labour market statistics produced. The process of deriving these population totals involves using the population counts from the most recent Census of Population as a base, and updating these each quarter using information on births, deaths and migration for that quarter. These population estimates are then revised once a new Census of Population has been completed. With the availability of results from the Census of Population 2016, the CSO has now revised the population estimates for the period Q to the current quarter using the new benchmark population totals for The introduction of the difference would be expected to be cumulative in nature, with relatively small differences in the earlier periods with the difference increasing over time. Given the critical importance of accurate labour market estimates, the CSO also revises labour market estimates using the updated population estimates. Therefore, from the publication of the Q LFS, the results published incorporate the new population estimates for each quarter since Q into the weighting methodology. The focus of the labour market estimates derived from the LFS is the working age population (persons aged 15 years or more). The original population estimate for the QNHS for the second quarter of 2016 for the working age population was 3,637,700 while the revised estimate from the Census of Population 2016 for the same quarter is 3,734,100 - a difference of 2.7% (96,400). 4.6 Backcasting of QNHS data series As a result of the move to the LFS, the previously published labour market series was revised. In order to minimise the breaks in the series to the key survey estimates, the CSO has created a backcasted QNHS series from Q to Q These backcasted series have been created using scaling factors from a comparison of data captured from a parallel run of the QNHS and LFS at the beginning of These backcast series are consistent with the LFS series that is being published for the following main indicators: Employed males aged years Employed males aged 25+ years Unemployed males aged years 13

14 Unemployed males aged 25+ years Inactive males aged years Inactive males aged 25+ years Males aged < 15 years Employed females aged years Employed females aged 25+ years Unemployed females aged years Unemployed females aged 25+ years Inactive females aged years Inactive females aged 25+ years Females aged < 15 years Scaling factors were also calculated for the NACE Rev. 2 groups published in Table 2a of the LFS release, broken down by sex, resulting in 30 additional categories. As a result, these series are also consistent with the new LFS series. It was not possible, however, to control for all of the series included in the QNHS release. Therefore, other than the headline indicators and NACE Rev. 2 series described above, there may be changes in the levels of other series from Q onwards and consequently users should exercise caution when comparing data from before and after this period. 4.7 Grossing and Weighting Once a final dataset has been created, the data is weighted to population totals. To derive grossing factors, population estimates are produced by the CSO s Demography Unit each quarter by sex, age (5 year age groups) and region (8 NUTS 3 regions). In addition, the Demography unit produce estimates of population for each quarter by sex, broad age (less than 15 and 15 or more) and nationality (5 nationality groups including Ireland, UK, EU13, EU13 (the EU excluding the EU15) and Rest of World). The individual returns are matched to these population estimates and the grossing factor is calculated by dividing the total estimates population in a given cell by the number of valid responses in that cell - i.e. if there are 1,000 respondents from the sample in a given cell and an estimated population of 40,000 then each of the respondents in this cell will have a grossing factor of 40. The grossing procedure is carried out using the SAS CALMAR statistical software routine which was developed by the French National Statistical Institute, INSEE. The CSO has introduced a non-response adjustment into the weighting procedure for the LFS from Q onwards. The adjustment applies extra weight to the groups who tend to be less likely to respond to the survey to make the results from the achieved sample more representative of the target sample and target population. 14

15 The grossing factors (now inclusive of the non-response adjustment) calculated for each individual cell are a function of the number of valid responses in that cell. It can be noted that the overall number of valid responses and the revised average grossing factor assigned in the LFS (formerly the QNHS) for recent quarters was as follows: Period Total valid responses Total valid responses persons age 15+ Q ,249 40, Q ,518 38, Q ,403 39, Q ,381 40, Q ,763 40, Q ,915 41, Q ,515 39, Q ,197 37, Q ,939 39, Q ,110 36, Q ,047 35, Q ,286 32, Q ,048 34, Q ,059 32, Q ,853 33, Q ,027 34, Q ,315 32, Q ,620 26, Q ,419 27, Average grossing factor assigned persons aged Computation of Outputs, Estimation Methods Used Output results are aggregated to produce the various totals published. These aggregations are usually produced using key variables such as sex, age region, ILO status etc. In general, all aggregations produced are done by way of various SAS procedures. The aggregate results produced for any given set of classifications will be the sum of the individual grossing factors of the valid responses which belong to that set of classifications and no estimates are made unless the data itself has been captured within the survey. 4.9 Other Quality Assurance Techniques Used A series of audits are carried out each quarter to verify the quality of the individual data being collected by interviewers during face to face interviews with respondents. 15

16 Interviewers receive regular feedback on the quality of the data they are producing and any unusual trends are highlighted and followed-up. Interviewers undergo training when first employed and regular meetings are held with field co-ordinators to ensure standards are maintained. Further training is also provided to interviewers where appropriate. Call centre interviewers also undergo training before commencing work on the LFS. Interactions between the call centre and survey respondents are audited on a random selection of cases. The audit process involves listening back to the telephone interaction and then rating the quality of the interviewer's work on a standardised score sheet. Interview outcomes such as complete or partial interviews and refusals are checked during the audits. When auditing an interview the questionnaire is reviewed to ensure that the data gathered was entered accurately. Any issues that arise are addressed initially with the call centre supervisor and subsequently at the quarterly training sessions where members of the CSO based staff deliver detailed training on any questionnaire changes. At an aggregate level, account is taken of other national economic indicators (e.g. Live Register trends, taxation trends) when quality checking the data produced from the LFS. 16

17 5 Quality 5.1 Relevance The data is required under EU Council Regulation No 577/98. The LFS is the primary source of a number of key national indicators including official estimates of employment and unemployment. In addition to the main labour market estimates published on a quarterly basis, LFS (formerly QNHS) data is used in a number of other reports and publications produced by the CSO. In the period between each Census of Population, the LFS provides estimates of various demographic and other social indicators such as levels of educational attainment. Given the critical nature of estimates produced the relevance of the LFS can be taken as very high. Other CSO sections which make use of LFS data include: National accounts Construction Earnings and employment Tourism External users who make significant use of LFS data include, but is not limited to: ESRI Central Bank SOLAS Health and Safety Authority The Oireachtas Department of Finance Department of Business, Jobs and Innovation Department of Education and Skills Department of Health and Children Department of Employment Affairs and Social Protection National media organisations 5.2 Accuracy and Reliability Sampling Effect & representativeness As the LFS is a sample survey it is subject to sampling error. Precision estimates are calculated using Jackknife replication for key variables. The table below shows estimated standard errors and confidence intervals for some of the key estimates for the LFS in respect of the most recent quarter. The 95% confidence intervals indicate the range within which we can be 95% confident the true value of the estimate in question will lie based on measurable sampling error. 17

18 Q Estimates of Standard Error, Coefficient of Variation and 95% Confidence Interval for key ILO results ILO Classification Actual measure (Thou) Standard Error (Thou) CV (%) 95% CI Lower Limit 95% CI Upper Limit (Thou) (Thou) Employed 2, , ,257.1 Unemployed 1, Total Labour Force 2, , ,398.7 Not in the Labour Force 1, , ,466.5 Q Estimates of Standard Error, Coefficient of Variation and 95% Confidence Interval for NACE Rev.2 Sector of Employment Sector of Employment Actual measure (Thou) Standard Error (Thou) CV (%) 95% CI Lower Limit (Thou) 95% CI Upper Limit (Thou) A Agriculture, forestry and fishing B E Industry F Construction G Wholesale and retail trade; repair of motor vehicles and motorcycles H Transportation and storage I Accommodation and food service activities J Information and Communication K-L Financial, insurance and real estate activities M Professional, scientific and technical activities N Administration and support service activities O Public administration and defence; compulsory social security P Education Q Human health and social work activities R-U Other NACE activities

19 Variance estimation for stratified household surveys is complex and there are a number of different methods. Further information is also available from the following links regarding variance analysis. It can also be noted that CSO continues to examine alternative methods of estimation and will expand the availability of measures for which such data is produced. %26p_p_state%3Dnormal%26p_p_mode%3Dview%26p_p_col_id%3Dcolumn- 2%26p_p_col_count%3D1%26_101_INSTANCE_8v4nUYMbAXCj_delta%3D20%26_ 101_INSTANCE_8v4nUYMbAXCj_keywords%3D%26_101_INSTANCE_8v4nUYMb AXCj_advancedSearch%3Dfalse%26_101_INSTANCE_8v4nUYMbAXCj_andOperator %3Dtrue%26p_r_p_ _resetCur%3Dfalse%26_101_INSTANCE_8v4nUYMbA XCj_cur%3D Non-Sampling Effects In addition to known sampling errors, any survey will be subject to other non-sampling errors (for example measurement errors arising from questions not capturing the desired information accurately). Non-sampling error is far more difficult to measure than sampling error and no formal estimate of non-sampling error is available in the LFS. Information on the interviews is collected and analysed to help minimise non-sampling effects (including, for example, when interviews were conducted and their duration). This information is compared across the interview teams to ensure no unusual variation in interviewer performance exists. Co-ordinators, as an additional check on the quality of the interviewer's work, call back to around 2% of households to check the quality of the collected data. Audit checks are also carried out on interviews carried out at the call centre Quality of the Data Sources used (other than survey register) Not applicable Register Coverage The entire stock of private households at the time of the most recent Census of Population in the country represents the full sampling frame for the LFS. The sample based on the 2011 Census was first introduced on a wave by wave basis in Q and was fully in effect as of Q Users should note that there is no overlap in sample between the QNHS and the LFS. 19

20 Non-response (Unit and Item) An adjustment for non-response was introduced into the weighting procedure for the LFS from Q onwards. The adjustment applies extra weight to the groups who tend to be less likely to respond to the survey to make the results from the achieved sample more representative of the target sample and target population. The table below gives a breakdown of the response and non-response beginning with the introduction of the new LFS in Q3 2017: Q Q Target households 27,300 28,600 Houses dropped* 6,911 6,998 Revised household target 20,389 21,602 of which: Vacant - CAPI 1,222 1,133 Uncontactable - CAPI 1,884 2,012 Uncontactable - CATI 1,222 1,538 Refusal CAPI 1,457 1,700 Refusal - CATI Other non-interview - CAPI Other non-interview - CATI Actual houses interviewed 12,989 13,436 of which: Insufficient data to assign grossing factor Final total number of houses used to generate outputs 12,961 13,419 * Households are sometimes dropped due to staffing shortages and unforeseen local circumstances. 20

21 For reference purposes, the table below gives a break-down of the response and nonresponse to the now defunct QNHS: Q Q Q Q Q Target households 26,000 26,000 26,000 26,000 26,000 Houses dropped* 3,560 4,338 3,960 3,859 4,921 Revised household target 22,440 21,662 22,040 22,141 21,079 of which: Vacant 2,377 2,478 2,458 2,391 2,337 Uncontactable 1,930 1,913 2,063 1,892 1,854 Refusal 1,687 1,623 1,691 1,734 1,594 Actual houses interviewed 16,446 15,648 15,828 16,124 15,294 of which: Insufficient data to assign grossing factor Final total number of houses used to generate outputs 16,420 15,632 15,808 16,116 15,287 * Households are sometimes dropped due to staffing shortages and unforeseen local circumstances Measurement Errors No formal evaluation of sources of error is available, although measures are in place to minimise error (details below). The quality of the data collected is improved using regular field staff training (including the use of video recording of training interviews) and debriefings for example, suggestions from field staff regarding the wording of certain questions.interactions between the call centre and survey respondents are audited 21

22 on a random selection of cases. The audit process involves listening back to the telephone interaction and then rating the quality of the interviewer's work on a standardised score sheet. Interview outcomes such as complete or partial interviews and refusals are checked during the audits. When auditing an interview the questionnaire is reviewed to ensure that the data gathered was entered accurately. Any issues that arise are addressed initially with the call centre supervisor and subsequently at the quarterly training sessions where members of the CSO based staff deliver detailed training on any questionnaire changes. Respondent effects - most of the requested information is readily available to respondents. Proxy responses are not allowed for certain questions (for example income). A lot of the national modules only allow direct responses to improve data quality. Comprehension errors - most of the terms used by the survey are readily understood, although some issues occasionally arise Processing Errors a. Data capture errors: These errors are minimised by logic checks and limits on values that can be keyed for each question in the electronic questionnaire at the data collection point. b. Coding error: Checks are in place to minimise this risk, particularly with respect to Industry and occupational coding. The coding is conducted in-house at the CSO using an automated coding facility which is reviewed by a small team of coding experts. This approach reduces subjectivity and coding error. Overall it increases the quality and standard of coding of these key variables Model-related Effects Not applicable. 5.3 Timeliness and Punctuality Provisional Results No provisional outputs are published Final Results Since the third quarter of 2012, results have been published in the ninth week after the end of the quarter (the end of quarter is the Sunday of the final reference week in the quarter). The target is T+56 days and actual timeliness for recent quarters is: Period End of quarter Publication date Timeliness (days) date Q /03/ /05/ Q /06/ /08/ * 22

23 Q /09/ /11/ Q /01/ /02/ Q /04/ /05/ Q /07/ /08/ Q /10/ /11/ Q /01/ /02/ Q /04/ /05/ Q /07/ /09/ ** Q /10/ /01/ ** Q /12/ /03/ ** * Note that the QNHS release was postponed as a result of a planned upgrade to the CSO website during the week of August 17 th to 24 th 2015 inclusive. ** Note that this delay arose due to the phased implementation of a major redevelopment of the CSO's household surveys in the context of a significant Household Survey Development (HSD) project. The main indicators which are published in the core release are ILO status (employed, unemployed etc.), industry of employment, occupation of employment, status of employment, duration of unemployment etc. Indicators are published by a range of classifications including sex, age, nationality, region of residence and highest level of education attained although not all indicators are published by each classification due to small cell sizes. The primary classification used for the LFS results is the ILO (International Labour Office) labour force classification. Labour Force Survey data on this basis have been published since The ILO classification distinguishes the following main subgroups of the population aged 15 or over: In Employment: Persons who worked in the week before the survey for one hour or more for payment or profit, including work on the family farm or business and all persons who had a job but were not at work because of illness, holidays etc. in the week. Unemployed: Persons aged who, in the week before the survey, were without work and available for work within the next two weeks, and had taken specific steps, in the preceding four weeks, to find work. Inactive Population (not in labour force): All other persons. The labour force comprises persons employed plus unemployed. Variables which are not published but required by Eurostat include for example work patterns, methods used in search for employment and current education involvement. Unpublished variables are often used in the calculation of European indicators such as life long learning and other education related matters. Such additional variables are provided by the Labour Market Analysis unit in ad-hoc data requests etc. 23

24 The LFS publications can be found on the CSO website Coherence Coherence checks are regularly carried out with auxiliary sources for example other CSO releases such as Live Register, Retail Sales and external sources such as taxation trends, redundancy information etc. In particular the trend of the Live Register going up or down is used to coherence check LFS results. 5.5 Comparability To ensure comparability with other official statistics, standard classifications are used for LFS estimates (e.g. NACE Rev 2 for industry, UK SOC 2010 for occupation). The classifications used are in the main set by EU regulation. The classification of industrial activity changed from NACE Rev 1.1 to NACE Rev 2. from Q The classification used for occupation coding changed from UK SOC 90 to UK SOC 2010 in Q In the main these changes were required due to changes in the applicable EU regulation. To facilitate users, the CSO backcast industrial coding data to Q and backcast occupation coding data to Q The LFS is undertaken on a calendar basis which was first adopted in the QNHS from Q as described under section 2.5 above and section below. All data from Q onwards was revised to calendar quarter data and results are available on the website. Again this ensures greater comparability with other quarterly data which is typically compiled on a calendar quarter basis (for example quarterly national accounts). A note on the changeover to NACE Rev. 2 and the introduction of calendar quarters is available on the CSO website in the labour market releases and publications section. Up to and including Q the annual population estimates have been calculated using the de facto definition of population (i.e. all persons present in the state). From Q onwards a new concept of usual residence was implemented, i.e. all persons usually resident and present in the state plus absent persons who are usually resident in Ireland but temporarily away from home and outside the state. This again ensures comparability with other demographic statistics where usual residence is taken as the most appropriate population definition. In 2008 revisions were produced to implement this change in methodology (see section 5.6.2). 5.6 Accessibility and Clarity Assistance to Users, Special Analyses 24

25 All publications are available on the CSO website. Information on methodology is also available at the following link The background notes on the publication provide some detail on the survey. For the core LFS publication, a press conference is held every quarter to enable users and commentators to fully understand the data or seek further clarification. Ad-hoc analysis can also be produced on request. Anonymised microdata (for all calendar quarters) is made available to researchers via the Irish Social Science Data Archive (ISSDA). Such data is accessible by researchers applying directly to the ISSDA. For further information see: Access to a Research Microdata Files (RMFs) can be requested from the CSO under the CSO s microdata access policy. Extensive use is made of this facility by the research community. For further information see: / Revisions The most recent revisions to the LFS data series were scheduled revisions arising as a result of the 2016 Census of Population. As a result of the Census, new population estimates were calculated on a quarterly basis using the 2016 Census of Population figures as a base. These new population estimates were then used as a new grossing frame for the quarterly LFS (formerly QNHS) data and as a result, the estimates previously published were revised to take account of these new grossing frames for each quarter. This work was completed in 2018 and results for the entire new set of data are issued with the publication of result for Q on 16 th January 2018 Inter-censal revisions of the type described above will be completed every 5 years after each Census of Population Publications Releases, Regular Publications LFS main labour market estimates are published every quarter with a target release data of T+56 days. In addition LFS quarterly releases, modules are published on an ad hoc basis. Further information regarding these modules is available from this link: Statistical Reports 25

26 LFS contributes data to a number of statistical releases in the office such as Men and Women in Ireland, Measuring Ireland s Progress, the CSO Yearbook etc Internet All LFS publications are available on the CSO website in publication format. In addition data is made available via the CSO s main databank dissemination tool and is also hosted on the CSO website in Excel format: Releases and publications Databank dissemination erly%20series/labour%20force%20survey%20quarterly%20series_statbank.asp?sp=l abour%20force%20survey%20quarterly%20series&planguage= Confidentiality The confidentiality of all information provided to the CSO by individual respondents is guaranteed by law under the 1993 Statistics Acts. All CSO office, field personnel and dedicated call centre personnel become "Officers of Statistics" on appointment and are liable to penalties under this Act if they divulge confidential information to any outside person or body. Extreme precautions are taken to ensure that there are no violations of this principle throughout the survey process. The laptops used by field staff in the data collection process are encrypted and contain several layers of password protection. Data collected each day is transferred to the CSO using a secure encrypted tunnel. There is a dedicated area in the call centre for LFS interviewing and access to this area is restricted those who have been appointed as Officers of Statistics for the LFS. There is a dedicated server at the call centre to administer the LFS and all computers used for the LFS at this centre are password protected. Each night, the data collected that day is collated into a single zip file, encrypted and password protected. The file is then uploaded to a secure Secure File Transfer Protocal (SFTP) site at the dedicated call centre. This CSO then extracts this zip file to a secure CSO location for processing. Data is only published in aggregate form and care is taken to ensure that the data are aggregated to avoid the indirect identification of respondents. Confidentiality is also ensured within the anonymised microdata by using coded variables instead of original values for key characteristics. For example, age groupings are provided instead of single year of age. 26

27 6 Additional documentation and publications The main CSO QNHS homepage can be found at the following link: Statcentral Ireland s portal to official statistics: Eurostat issue many releases which use LFS data and the central repository for such data can be found as below: 27